Daily Intelligence: AI No Longer Asks Permission, It Asks for Power

Today's thesis, Thursday, July 23, 2026, is uncomfortable but useful: artificial intelligence can no longer be understood only by looking at models, demos or product presentations. It has to be understood through energy, oil, debt, memory, subsea cables, permits, regulation and corporate results. AI no longer asks permission to enter the economy. It asks for power. And when something asks for power at this scale, it stops looking like an app and starts looking like critical infrastructure.

I did not find today's Bible verse, Glorify reflection, Stoic reflection or app of the day in the allowed memory files. Still, the human thread of the day is clear: it is worth separating enthusiasm from endurance. Enthusiasm starts a wave; endurance is proven when financing costs rise, oil tightens, regulators appear and users decide whether a promise actually improves their lives.

Macro / Energy

The first data point that matters today does not come from a lab, but from the oil market. AP described yesterday's session as one where Wall Street traded almost flat while oil rose roughly 3%, pushed by tension linked to Iran and fears of disruption around key energy routes. The same pressure left U.S. crude above 94 dollars and put inflation back on the table just as investors wanted to talk about earnings, chips and productivity.

That detail matters because the AI cycle depends on a macro condition that is often forgotten: abundant capital and reasonably predictable energy. If energy prices start contaminating inflation expectations again, central banks have less room to sound patient. If rates stay high, the cost of capital rises. And if the cost of capital rises, building data centers, reserving electricity, financing GPUs and sustaining demanding valuations becomes harder.

The British reading adds nuance. Market coverage yesterday noted that U.K. inflation fell to 2.6% in June, helped by softer food and diesel prices. Under normal conditions, that would be a clean signal of relief. But conditions are not normal: crude is moving sharply again because of geopolitics, and AI infrastructure is lifting electricity demand at the same time many countries are trying to electrify transport, industry and cooling.

The macro conclusion is not that AI is bad for the economy. It is more precise: AI is exposing a physical constraint. For years, software seemed to live in an almost abstract cloud. Now that cloud needs turbines, substations, liquid cooling, long-term contracts and grids capable of absorbing demand spikes. Xataka captured this technical shift this week through Nvidia's energy-efficiency framing: efficiency is no longer only about consuming less, but about completing more computational work per kilowatt-hour. That line marks the change in the era.

For markets, energy is today's reality detector. If crude stabilizes and inflation does not reheat, the productivity story can stay alive. If oil climbs and electricity becomes more expensive, AI will have to prove returns sooner, because the world will be less willing to accept a rising bill in exchange for vague promises.

Geopolitics

Today's geopolitics has an obvious layer and an underwater one. The obvious layer is the Middle East, oil and sea routes. When markets start watching Hormuz, Bab el-Mandeb or U.S. gasoline prices again, the message is simple: the digital economy still depends on vulnerable physical points. There is no cloud without ships, energy, minerals, gas, cables and insurance.

The underwater layer is visible in Asia. Xataka covered China's progress in antisubmarine capabilities, including an improved Y-9 variant and a broader network of sensors, buoys and maritime surveillance. At first glance, it looks like a military story separate from technology. In reality, it belongs to the same map: controlling the sea means protecting routes for energy, data, chips and trade. Subsea cables and semiconductor logistics are as strategic as the ships around them.

The technological-sovereignty front is also widening. U.S.-China tension is not limited to advanced chips. It extends to AI models, data, standards, exports and access to sensitive capabilities. Reuters and Xataka have been pointing in that direction: if Washington can restrict access to advanced technologies on national-security grounds, Beijing is studying its own tools. The AI race is entering a less innocent phase. It is no longer only about who has the best model; it is about who decides who can use it, train it, export it or integrate it into critical sectors.

Europe is in an uncomfortable position. It has regulatory strength, consumers, talent and a more demanding public conversation around privacy. But it depends too much on U.S. clouds, Asian semiconductors and exposed energy chains. It can set important rules, but if the infrastructure is built elsewhere, its power has a ceiling. Europe's question for the coming months is whether it can turn regulation into a trust advantage or whether it ends up regulating a race others are running with more capital and more available power.

Even Microsoft's decision to bring original Xbox games to Windows through official emulation fits this block. It looks like entertainment, but it speaks to a bigger issue: who preserves access to digital goods and under what conditions. Software preservation, account dependence, licenses, stores and platform continuity are part of everyday technological sovereignty. Not everything is national defense; some of it is cultural memory, digital ownership and control of access.

AI / Tech

Technology enters the session with markets focused on earnings and AI spending. Alphabet, Tesla and Intel concentrate attention because they represent three different questions. Alphabet must show whether data-center spending translates into cloud, advertising, internal productivity and competitive defense. Tesla must separate execution from narratives around autonomy, robots and energy. Intel must show whether it can find a place in a semiconductor map dominated by memory, accelerators and advanced manufacturing.

The pressure on Alphabet is especially clear. In a calm market, investors can accept years of investment in exchange for a credible promise. In a market with high oil and sensitive rates, they ask about returns sooner. Every dollar of AI capex needs an explanation: more cloud revenue, better advertising margins, subscription products, operating savings or a defensive barrier against rivals. If the answer is only 'we have to invest because everyone is investing,' the market becomes less patient.

The other layer is hardware. A rebound in semiconductor and memory names does not erase the underlying fragility. AI has restored glamour to components most users never see: HBM, DRAM, NAND, low-latency networking, cooling systems, cabling, power supplies and orchestration software. On good days, those bottlenecks create extraordinary profits. On bad days, they remind investors that margin can migrate within the value chain. The winner is not always the company with the prettiest app; sometimes it is the company controlling the scarce part.

Xataka's recent article on Nvidia and liquid cooling lands this point well. The battle is no longer just about making more powerful chips. It is about extracting more inference and training from each unit of energy, keeping temperatures under control, designing denser racks and preventing the data center from becoming an unmanageable thermal factory. The most profitable innovation may not be in a shiny interface, but in an invisible efficiency gain.

There is also a cultural reading. Microsoft emulating Xbox on Windows reminds us that technology platforms are beginning to assume archival responsibilities. The digital industry has spent years selling access, not ownership. Now users are starting to ask what happens to a library when a store changes, a service closes or a license expires. The same question applies to AI tools, models and business data: what part of the value do you control, and what part lives inside a platform that can change the terms?

Markets

The market is trying to do two things at once: sustain the bullish AI story and price an energy shock. AP pointed to a mixed session, with the S&P 500 almost flat, the Nasdaq weaker and sharp moves in individual names. Super Micro jumped on better margin forecasts, while Alphabet slipped before results that investors will scrutinize for AI investment discipline. That divergence tells the story: the market is not buying everything technological; it is selecting what it believes can turn demand into profit.

Oil is the silent rival of valuations. A barrel near the mid-90 dollar area does not kill a bull market by itself, but it changes the tone. It lifts logistics costs, pressures inflation expectations, complicates central-bank communication and lowers tolerance for businesses promising profits far in the future. In other words, it makes the future worth a little less if the present becomes more expensive.

Fixed income matters too. The move in the 10-year Treasury yield toward the 4.65% area, according to AP's coverage, signals that the market is not relaxed. For AI companies with strong balance sheets, that is manageable. For leveraged projects, debt-financed data centers or smaller suppliers dependent on constant funding rounds, it is more delicate. Digital infrastructure is financed like infrastructure: with maturities, collateral, supply contracts, customer commitments and rate sensitivity.

For practical investors, the useful separation remains threefold. First, direct AI exposure: semiconductors, cloud, software and platforms. Second, infrastructure exposure: energy, grids, cooling, land, construction, financing and security. Third, regulatory exposure: social platforms, digital identity, privacy, content, copyright and compliance. The first offers more beta. The second can offer more contracted revenue. The third can either protect or punish business models depending on how rules evolve.

Today's temptation is to chase any rebound in chips. The more serious reading is different: look for who wins even if energy costs rise, who can pass costs through, who has the balance sheet to endure, who converts AI into measurable productivity and who depends only on enthusiasm returning. The difference between those categories will matter more if the market moves into a phase of selection rather than broad euphoria.

24-72h Radar

First, megacap earnings. Alphabet and Tesla are the immediate thermometers. For Alphabet, the market will watch capex, cloud, margins, AI integration in search and advertising, and any sign of discipline. For Tesla, it will watch deliveries, margins, energy, autonomy and whether the robot and robotaxi narrative rests on verifiable milestones.

Second, oil and the Middle East. If crude stays high or climbs again, markets will have less room to ignore inflation. If it cools quickly, earnings and guidance will regain the spotlight.

Third, semiconductors and memory. The recovery will be more credible if it comes with orders, firm pricing, controlled inventories and real end demand. If it is only short covering, it may not last.

Fourth, AI energy infrastructure. Watch power-supply contracts, utility agreements, data-center tariff regulation and any signal that consumers or politicians are beginning to resist paying for grid expansion.

Fifth, digital sovereignty. China, the United States and Europe will keep moving around chips, models, data and access control. The next measures will not always arrive as major announcements; sometimes they will appear as technical standards, licenses, security reviews or localization requirements.

Scenario Conclusion

Base case: oil remains uncomfortable but does not break higher, megacap earnings show that AI spending still has commercial logic and semiconductors rebound selectively. Practical implication: keep exposure to quality infrastructure, software with verifiable returns, cybersecurity and flexible energy, while avoiding any company that only carries AI in the label.

Bull case: energy tension falls, Alphabet convinces on monetization and capex discipline, Tesla shows more execution than promise and memory suppliers confirm solid demand. Practical implication: increase risk gradually in companies with visible orders, strong balance sheets and the ability to turn AI into cash, not just headlines.

Bear case: crude rises again, bond yields apply pressure, markets interpret AI spending as excess and regulators further fragment operating costs for platforms. Practical implication: reduce technology beta, prioritize liquidity, resilient balance sheets, contracted revenue and essential infrastructure over distant-growth narratives.

The story of the day is not that AI has lost its shine. It is that it has gained weight. It no longer floats above the economy as an abstract promise; it lands on power grids, balance sheets, ports, cables, rules and bills. That makes it more real, not less important. But when a technology becomes real, the market changes the question: it stops asking only what it can imagine and starts asking how much it can bear.